Multiple regressions for assessing the effects of geographic and ecologic 

 variables on the hardiness of populations were made separately for data from two 

 phases of cold acclimation (table 1) . Whereas the regression for phase one accounted 

 for a statistically significant (1 percent level) 38 percent of the variance, that 

 for phase two accounted for a nonsignificant 15 percent. In accordance with relatively 

 little variance among populations, variation in hardiness during the second phase of 

 cold acclimation could not be related to geographic or ecologic variables of the seed 

 source. 



Table 1 . --Results of multiple regress-ion analyses of hardiness during phases 



one and two of cold acclimation on geographic and ecologic variables. 

 Variables are defined in the text. Results are presented in terms of 

 regression coefficients (h) and standardized partial regression 

 coefficients (b') 



Hardiness during: 



Phase one 



Phase two 



Variable 



b 



Latitude 



Longitude X2 



Elevation J3 



Habitat type: 



Pseudotsuga menziesii 

 Abies grandis Z5 

 Thuja ocoidentalis 

 Tsuga heterophylla Xy 



-9.27 

 2.03 

 -3.05 



.96 



-4.66 

 -.62 

 -2.09 



■0.59 

 .08 

 -.37 



-.02 

 -.13 

 -.02 

 -.04 



0.51 

 -2.96 

 .82 



14.74 

 21.60 

 20.92 

 11 .91 



0.03 

 -.12 

 .11 



.41 

 .64 

 .64 

 .28 



bo 

 i?2 



yx 



111.40 

 .38 



14.35 



29.26 



.15 

 15.18 



By contrast, a comparison of standardized partial regression coefficients 

 reflects the relative importance of the continuous independent variables in predicting 

 hardiness during phase one. Elevation and latitude control differentiation 

 (table 1) . Longitude has little influence. Since regression coefficients for the 

 constant terms (bi^ to bj) are deviations from the intercept Zjq, under the present 

 model the constants reflect deviations in mean hardiness of seedlings representing 

 the indicated habitat types from that of seedlings from the Abies lasiocarpa habitat 

 types. None of the coefficients for the constant terms are statistically significant. 



8 



